Lose the Loaded Dice



  • Look, your 50/50 chance of extra casualties. . .fucking sucks to be absolutely frank.

    I understand why it is a weighted mechanic. It makes up for the clueless AI and makes games with more than one player more competitive. . .But destroys single player games. I cannot win a game anymore without save scumming because every AI gets their extra casualties every time they attack, while I get it maybe a third of the time, and they also proctor every time they defend, where as I might see it a quarter of the time on defense. I can't consistently take double damage and deal half damage and still be effective in a game, and I don't think many others could either. Yes I have the setting on hard, yes I choose underdog nations. Yes I am playing the largest and most complex maps available like World at War and Liebowski variants. I accept not every battle is going to go my way and this has always been a present issue with Low Luck games. . .So I used to avoid it by simply not using low luck. . .but as of the last few builds, that doesn't make a difference. I have no idea what has changed, but clicking off no luck doesn't result in more randomized results as it used to.

    I have been playing this off and on for a decade, its my background activity when I am waiting for things to render or stack and have to keep an eye on things to ensure nothing fails out. . .And man this wrecking my joy.

    I don't want to muck things up for anyone else. I am not asking for a sweeping change to how dice are used elsewhere, just please consider adding a setting specifically for single player games where the extra damage can be turned off completely or be used with less weight added for the AI.

    If you think I am delusional, please, play World at War with low luck and without as Germany with everyone else hard. See if you can win a game without having to either take control of an allied AI, or save scumming the major battles. I am on the level with this.


  • Admin

    TripleA uses a library from "apache" for random number generators. The random number generator is the mersenne twister algorithm that they have implemented.

    More information on Mersenne Twister can be found at:
    http://www.math.sci.hiroshima-u.ac.jp/m-mat/MT/emt.html

    Any random number generated used by a computer will never be truly random, but should generally be random enough that outside of a casino the next number is very difficult to predict and 'random enough'.

    The code that generates a dice roll is here:
    https://github.com/triplea-game/triplea/blob/master/game-core/src/main/java/games/strategy/engine/random/PlainRandomSource.java#L35

    You can observe we simply invoke the API of the random number generator "getNext(max)"

    The dice roller used by the AI and human player is the same. The game is implemented such that the only difference between two players, AI or human, is whom to ask for decisions. When a decision needs to be made to move units for example, the AI computes and gives a sequence of movements, whereas a human player uses the UI to generate the same movement data. Otherwise when it comes to battles, there is nothing special about an AI or a human, both use the same logic for getting and generating random dice rolls.

    For healthy skepticism, I can check that our usages of bounds is correct. If that were off, then it would be off uniformly for all players (AI included). I am skeptical that is the case, but what is certain is the code does not have a special roller for AI.

    The code to trace that is super convoluted, disappointingly so, but we can go through the exercise of tracing further how the random number goes on to be everyones dice rolls.


  • Moderators

    @anEshva i don't believe the dice are "loaded" from my own observations.



  • Maybe I am neurotic, but I really don't think so.

    I just resent having to cheat to win, it makes victory meaningless -all the moreso when one is playing alone- and over the last three months or so it seems as though my outcomes have been different than those I have come to know over the years.

    You say its honest then I won't call you a liar. . .But I hope you will take me up on my proposal. Just play two games of WaW, one with low luck and one without and see if it doesn't feel as though a disparity exists between how often the additional casualties proc for AI and player. I fully expect to lose when playing Italy, or Finland (the latter being my favorite, because I am a sucker for punishment), but with Germany it should be a coin flip as to whether you can beat Russia down sufficiently before the US comes swinging through. . .

    Admittedly, it could also be the way I play. With Germany I tend to build two battlleships first turn and keep a balance between naval and ground forces to keep the US out of the Atlantic for a couple more turns as I can contain the British. It really could be my margins are so tight even just a minor variation is enough to offset my chances. . .Except I have been playing for so long with more or less the same strategies that I cannot convince myself that is the case.

    Anyway I appreciate you getting back to me and will leave this in your capable hands.


  • Admin

    I love the passion. From a coding perspective, there is no bias between players, they all grab from the same random number generator pool. I'll do some research in the background to really confirm the numbers are not skewing high (I think that's extremely unlikely to be the case, but it's worthwhile to improve the dice rolling code and really get from 99.9% certainty to 100%).


  • Admin

    When it comes to dice rolls and battles, I do think in general two factors really come into play:

    • confirmation bias
    • conditional probabilities

    For the first one, we really remember those battles that we need, meanwhile the other eccentric rolls that don't matter, or all of the 'normal' rolls just do not stick out in our mind.

    The conditional probability case is an interesting one. In some large battles, it's very interesting that with 1 units less, it's 100% for defender, add one attacking unit and you get 50%, and then add one more and it flips and is 100%. The margins are amazingly close.

    The 100% is I don't think actually is accurate. Regardless, for battles that are 95%, one thing that can be important is just the first round. Basically, in 5/6 battles you're going to win. If you go down that path, then all combinations of dice rolls leads to wins. But perhaps in 1/6 battles, you wind up in that "one unit less case", and suddenly are you doing a 40% or a 50% battle. One side of the probability tree (5/6) is almost all victories, the other side (1/6) is mostly defeats.

    So when someone cries "what, I lost a 95% battle!", sometimes I wonder, yeah, you lost a 40% battle given the AA hit their 1/6 chance, so it's not that surprising at all actually.

    In another example, if the attacker misses every extra dice roll, and the defender hits, then the same conditional probably can kick in, a 95% battle becomes 70%, becomes 40%, becomes certain defeat.


  • Admin

    @anEshva re: vs AI on WaW

    I'm really surprised you play axis vs an allied Ai. The reason for the surprise is the AI is not quite good enough to grasp the complexity of long supply chains, particularly naval. I have been surprised at how reasonably well the AI does do though.

    I just played vs AI with LL. The rolling of the dice did not really seem to matter. I made a few pretty bad blunders but in the end was at a cake walk by round 6. The US spent the whole game conquering South America and building an unbelievable number of big transports. Russia was pushed back and was too aggressive against Germany and didn't put up too much of a defense. Leningrad fell, the UK was not effective to harrass western europe and the axis pushed in. In the east Yamamoto ran wild in the pacific and hisaichi over-extended in china but not after sticking a dagger quite deep.

    Happily WaW is my favorite map. Playing it through here I can't help but see various features and game items for betters/smoother mechanics and ways to play the game faster. My personal wish list is quite long for what I would like for the game, meanwhile we are saddled with some very important technical projects that need to get done (map uploads, new network technology that will remove the bots in favor of 'network bridges', text-based save games that will improve compatibility and flexibility when coding). Hopefully when that list is done we can dig into some more interesting features. FWIW, on that list is a dice server as we need to migrate MARTI to not be PHP.

    @anEshva I do think perhaps that the 2bship german build might be the problem more than dice ; )
    G in the first 2 rounds needs to ensure it'll capture paris, a few transports can be purchased but large navy is not needed and bad cost-benefit that early in the game. UK takes a while before it can start to threaten the northern sea, and even then German air can keep them at bay while transports do their work to increase mobility. I hope that the bad dice rolls haven't soured you too much, there is a lot of strat to work on in WaW, cheers.

    waw_as_axis-vs-ai-victory-round6.tsvg



  • @LaFayette

    Alright, well, sorry for ringing the alarm then. Guess I just lost my touch. At least I have some uphill battles to look forward to getting it back.

    I can't play as the Allies. It's a little too easy. As the Axis, all the positions are kind of fun because no matter how well you do yourself someone else is folding up elsewhere and you are constantly pivoting to compensate and alleviate the pressure on them. I like that. It is also kind of rewarding when something totally unexpected happens like Taiwan taking Borneo then surging through India. Its not the same watching China push Hisaichi's shit in

    Trust me, I also have issues with the AI, but if I play more than one country the game becomes a steamroll. Up until about round 8 where I usually take Russia the AI is competent enough. After that Japan starts doing some really wonky stuff and becomes really unreliable, if Italy hasn't managed to secure north Africa at that point then the US will be a nuisance, and even if they have they also begin doing some really wonky stuff. . .And you are absolutely right a lot of it has to do with sea warfare.

    The Axis has tremendous naval issues but the US fleet seems to work pretty well. Its far more aggressive and effective than Japan. I really don't know. The British are very decent also. Not sure where the problem could be.

    If I have one major grievance it is Romania, who would be best served taking Greece along side Italy and opening the Turkish front hopefully with Italian support again. I would also prefer if all the AI would stick to their own fronts rather than constantly trying to reinforce my own. It is exasperating taking a territory, losing it the next turn to Russia, only to have another country take it and force me to defend their claim in order to continue to push.

    Lastly, I am well aware of confirmation bias, but the game prior to my posting I had been tracking the rolls with some scratch paper and made a general estimate of the odds that I ended up with, and they felt about on par with what I had been seeing. I wouldn't have said anything if I had been wrong. I had been biting my tongue on this for weeks. I still maintain losing the option of extra casualties would improve the single player experience though, without the chaos factor is comes down to pure strategy which, well, does short the AI a bit but in a single player run that hinders as much as it helps since much of one's success is contingent upon your allies.

    EDIT: Yeah, I have never taken Paris in two rounds in WaW. Round 4 is the norm, but at that point I usually have naval superiority over the Atlantic which allows me to sweep through Spain as I take Russia, I find myself in good position at that point to sack Britain and hit the US properly. . .Not that that has been working for me lately. lol

    It bares mentioning I have never actually used the battle simulation tool to figure out my odds of success and frequently do ill advised things like using bombers to mop up isolated units behind enemy lines to break up reinforcement chains. Nothing is more bitter than losing a bomber to infantry.


  • Admin

    @anEshva said in Lose the Loaded Dice:

    I still maintain losing the option of extra casualties would improve the single player experience though,

    Could you elaborate a bit on what you mean by 'extra casualties'? I just want to be sure we would be talking about the same thing.

    Yeah, I have never taken Paris in two rounds in WaW. Round 4

    Indeed, AFAIK G2 Paris is a legend, I have hear it can be done but it involves using italian aircraft to strafe and an awkward purchase of mech infantry from germany.

    G4 Paris though seems late! Round 3 is very do-able, it involves pushing most tanks towards france and first round buying a mix of 7 cars and/or tanks. Then in the 2nd round, Germany should buy a few stukas and 2-3 mech infantry to ensure that France can't secure the odds by purchasing all tanks in Paris.

    Anyways, I digress...


    I did look at the roller logic a bit more, there is no rounding happening, any deviations are going to be problems in the mersenne twister library we are using. I think that's going to be surprising if there were problems there. With a random number generator and saying how likely things are, it's very difficult. Are the dice loaded? Or was that a 1% game with odd dice? After what point can a person actually say dice are loaded..



  • There is a concept in statistical analysis which I cannot remember the name of, the something valley, which explore those ends of the spectrum whom experience the rare phenomena of persistent aberration, sequential streaks of good or bad luck. Its a known thing and all one can do is shrug it off. Assuming of course I am not delusional or a just generally full of shit, I could have just been seeing statistically unlikely results and then tied them the recent spate of updates conflating correlation and cause. Its a pretty human mistake.

    Yeah turn three is usually pretty solid even without any investment. Just send all infantry and aircraft east except what directly borders France and Belgium, and all tanks west (except the four in reserve from Silesia and Pomerania used to reinforce E Prussia). With that you can pretty much waffle stop both France and Russia without any effort. I usually wait the extra turn on France to get Britain to reinforce the capital with its aircraft, which makes clearing the Atlantic almost as easy as sacking France. Also, the more I send in, the more I retain after, and I like to take Spain fairly quickly. In a multiplayer game I would be less concerned about any of that, but since none of the Axis AI seem to be able to do anything meaningful at sea and Italy won't bother with more than or two Spanish territories it sort of falls on me to exploit those

    As for the extra casualties, I just mean the 50/50 die



  • And yeah, I swapped out a battleship with a couple tactical bombers and am back to my expected results.

    Lets just chalk this one up to me being an idiot and never speak of it again 😜

    X.tsvg


  • Moderators

    @anEshva You speak of the 50/50 die. As far as I know there is no such thing in TripleA. I suspect you play Low luck. Sorry if I misunderstand and you know all this.
    If you attack with 7 infantry 1 hit is assured. The remaining roll is for 1 (1/6 chance), not a 50/50. (6+1=7)
    4 defending inf, remaining roll is 2 (6+2)
    3 armour, remaining roll is indeed 50/50 (6+3)
    11 attacking inf, remaining roll is 5 (6+5)

    Perhaps the AI is making good use of this to bring enough units to make it's remaining rolls mostly over 50%.


  • Admin

    @Zjelcop said in Lose the Loaded Dice:

    Perhaps the AI is making good use of this to bring enough units to make it's remaining rolls mostly over 50%.

    IIRC it generally is. It will favor battles that are around 70% or better odds.


Log in to reply